The April 2025 ERCOT Price Spike

The morning of April 7, 2025 began unremarkably. It was a Monday in early April, the slack season on the Texas grid, when demand sits well below its summer peaks and the weather is mild. There was no storm on the coast, no Arctic air over the plains, no heat wave, no emergency alert, no conservation appeal. By every measure that usually moves an electricity market, April 7 looked routine. Then, for about ninety minutes before dawn, the real-time price in the largest power market in the country rose to levels normally associated with grid stress.

The grid was not in trouble. That is the notable feature of this event. The three other ERCOT events in this series each have a clear cause: a four-day freeze that nearly took the interconnection down, a hurricane that darkened a city, a total eclipse that blocked the sun. April 7 has none of that. Load was unremarkable, the air was cool and still, and the lights stayed on. The event was entirely a price, and the price was set by three conditions lining up at the same hour: a dispatchable fleet thinned by spring maintenance, a wind forecast that came in 2.4 gigawatts too high, and a windless dawn before solar had come online. None of those is unusual on its own. Together, for ninety minutes, they produced a price of more than four thousand dollars a megawatt-hour.

Press play and scrub across the morning of April 7. There is little to see on the map the way the other stories show something: no cloud mass, no shadow, no darkened zone. That absence is the point. This was an event that happened entirely in the market, in the five-minute prices, while the physical grid stayed calm. The camera holds on the whole state because the spike was statewide, and the way to follow it is in the numbers below.

Start with the price itself, because everything else follows from it. The ERCOT-wide average hub, HB_HUBAVG, is the number traders watch as a read on the whole market. Through the small hours of April 7 it was quiet: around twenty-five to a few hundred dollars a megawatt-hour, the ordinary churn of an overnight market. Then, over a few five-minute intervals around 6 a.m. Central, it rose sharply. The average-hub price crossed $1,000 at 6:35, peaked near $3,985 in the five-minute bucket at 6:50, and reached a single-interval high of $4,347 at 6:54 (ercot.lmp_by_settlement_point). Within ten minutes it was falling. By 7 a.m. it was back to about $482, and by 8 a.m. it was near $25. The whole excursion, from calm to nearly four and a half thousand dollars and back to calm, fit inside about ninety minutes.

That shape, a steep rise and an equally steep collapse, is the signature of a scarcity event, not a sustained shortage. A grid that has genuinely run short of power stays expensive for hours; that was Winter Storm Uri, pinned at its $9,000 ceiling for days. A grid that is short for a single ramp prices that one moment and then relaxes once the missing megawatts arrive. April 7 was the second kind. Something was briefly scarce around 6 a.m., and then it was not.

The first question to ask of any price spike is whether it was systemwide or local. Texas is a big grid, and a single congested transmission line can drive the price at one hub sharply higher while the rest of the state pays normal rates. That is congestion, not scarcity, and it points to a constraint on the wires, not the supply. So look at the three major trading hubs separately. If only one moved, this was a local transmission constraint. If all three moved together, the whole system was short.

They moved together. At the peak of the spike the Houston, North, and West hubs were within about eighty dollars of each other, and their single-interval highs land close together: $4,605 at Houston, $4,566 at North, and $4,524 at West, all at 6:54 a.m. Central (ercot.lmp_by_settlement_point). Three prices, hundreds of miles apart, peaking in the same five-minute interval within a couple of percent of each other. That is the signature of systemwide scarcity. There was no cheap power anywhere in Texas to undercut the expensive power, so every part of the state paid nearly the same high price at the same time.

The clean-energy forecasting firm Amperon published the closest thing to a contemporaneous post-mortem of the morning, an analysis by Andrew Cipolla three days later, and its account lines up with the data point for point. Amperon recorded five-minute real-time prices climbing above $4,000 per megawatt-hour at roughly 6:50 a.m. Central, the highest ERCOT had seen since December 2024, and traced the cause to a supply shortage at the dawn ramp rather than to any single failure. The piece is the basis for what follows.

Here is the core of the story, the number that ties the price spike to forecasting. ERCOT runs two markets stacked on top of each other. The day-ahead market clears the afternoon before, when participants and the operator commit generation against a forecast of tomorrow. The real-time market then settles what actually happens, five minutes at a time. When the two agree, the forecast was good. When they diverge, the gap is a measure of the surprise. On April 7 the gap was large.

For the 6 a.m. hour, the day-ahead market had cleared the average hub at $53 per megawatt-hour. Real time that same hour averaged $1,821 (ercot.spp_day_ahead_hourly, ercot.lmp_by_settlement_point). That is a difference of roughly $1,768, a gap of about thirty-four to one between what the market expected and what the morning delivered. And the day-ahead market never came close all day: its highest hourly clear, at any hour of April 7, was just $96. The afternoon before, with all its models and all its committed generation, the market had priced the dawn as ordinary. It was not, and that was not apparent until it arrived.

Amperon located the mechanism in that gap. ERCOT's own day-ahead forecast for the morning, the firm found, sat 2.4 gigawatts above what the grid actually faced, and a miss that size meant the dispatchable plants that could have eased the ramp were never told to be ready for it. Generation that needs an hour to come up cannot help if the system only recognizes it is short fifteen minutes before dawn.

If the day-ahead miss explains why the grid had no warning, two other conditions explain why being short at that particular hour was so expensive. The first is the calendar. April is spring maintenance season on the ERCOT grid, the slack months when demand is low and generators take their plants offline for scheduled repairs to be ready for summer. Amperon put total generation outages on April 7 at roughly 36.6 gigawatts, high precisely because so many units were down for maintenance. The fleet that would normally have deep, cheap reserves to draw on at dawn was, that week, running thin.

The second condition is the wind, and it is the reason the West Texas turbines in these photographs matter. Wind is one of the largest sources of power on the Texas grid, and overnight it usually carries a substantial share of the load while the sun is down. On the morning of April 7 it produced little. System-wide wind output fell to roughly 3,082 megawatts at its trough around 3 a.m. Central and was still only about 4,500 megawatts during the spike hour (ercot.wind_generation_geo), a fraction of the fleet's multi-gigawatt capacity. The wind farms were not broken. The wind simply was not blowing, on a calm clear night, in the window before solar could take over.

Put the three together and the picture is clear. A thin fleet meant shallow reserves. A wind forecast that missed by 2.4 gigawatts meant the reserves that did exist got no notice. And a windless dawn meant that in the hour before solar generation began, there was a physical gap of several gigawatts between the supply the grid was counting on and the supply it had. Around 6 a.m., for a short window, ERCOT was short of cheap power, and it had to draw on the most expensive resource available.

When the cheap power runs out, the price is set by the last and most expensive resource the grid has to call on. On April 7 that resource was batteries. Amperon found that around hour-ending 8, with no cheaper generation left, ERCOT drew roughly 2.65 gigawatts from battery storage to fill the gap. A battery is not a power plant with a fuel cost; it is a finite amount of stored energy whose owner decides what that energy is worth. With nothing cheaper available to undercut them, the batteries could set their price, and that price was high. That is what set the spike. Not a failure, not a shortage of installed capacity, but the marginal megawatt being a battery that was the last resource available.

This is also why the collapse was as fast as the spike. Batteries set the price only for as long as they were the marginal resource. Once the slower dispatchable plants finished ramping up and the first solar came online, there was again something cheaper than a battery on the margin, and the price fell back to it quickly. Amperon described the same mechanism in reverse: cheaper generators finished ramping, ERCOT could wind down the expensive battery dispatch, and prices returned to stable levels. The expensive resource was only filling a ninety-minute gap, and when the gap closed, so did the price.

It is worth being precise about what this was not, because the instinct when a price hits $4,000 is to assume the grid was straining under high demand. It was not. At the moment of the spike, system load was only about 45.6 gigawatts (ercot.system_load), an ordinary spring-morning figure, well below the system's limits. Load did climb to a daily peak near 52.5 gigawatts, but not until early afternoon, hours after the price had already fallen back to $25. The spike and the day's high demand did not occur at the same time. This was a scarcity of supply at a thin hour, not a surge of demand against a ceiling.

The weather record rules out the other usual suspect. There was no heat and no cold to blame. Historical observations for Dallas Love Field show the morning was calm and clear, about 40 degrees Fahrenheit with zero-mile-per-hour winds at 6:53 a.m., the minute the price peaked, and the day topped out at only about 67 degrees (open_meteo.era5_grid records the same mild footprint over the state). The zero-wind reading is the central detail of the event: it is both why the turbines sat idle and why there was no storm to point at. The dawn was calm, and that stillness is what made it expensive.

ERCOT itself did not treat April 7 as an event. The operator's own April 2025 Monthly report makes no mention of the price spike, because from a reliability standpoint there was nothing to report: the grid never approached an emergency even as prices briefly cleared $4,500. The same report shows how routine the season looked from inside the control room. It noted that the system had just set an all-time March demand record of 62,920 megawatts the month before, and it projected a less than one percent chance of declaring an energy emergency alert in May. Comfortable conditions, in the operator's view, with isolated morning price events that did not register as risks to reliability.

Step back and April 7 stands out further, because the first half of 2025 was, by ERCOT standards, calm. The analytics firm Tyba Energy, surveying battery performance across that stretch, found that average day-ahead energy prices across nodes exceeded $500 per megawatt-hour in only about four intervals in the entire first six months of the year. High prices were rare. Against that backdrop, a real-time surge to $4,347 is an outlier of the kind the market produces a handful of times a year, and it came in real time at dawn rather than in the day-ahead market most participants watch.

For a sense of scale, Tyba separately broke down a later ERCOT spike, an evening event on April 20, 2025, that peaked near $400. That was a documentable price excursion worth writing up. The April 7 dawn event was more than ten times larger. Two spikes two weeks apart, with one about an order of magnitude above the other, is a fair measure of how far outside the ordinary the morning of the 7th was.

And the timing of it, dawn rather than a hot summer afternoon, was a preview rather than an aberration. Analysts spent the rest of 2025 documenting the same pattern: as solar reshapes the net-demand curve, ERCOT's price spikes have stopped confining themselves to August heat and increasingly occur at the edges of the day, the pre-dawn ramp before the panels come online and the evening drop after they fade. April 7 was an early, clear instance of that. The cause was not the weather. It was the shape of a grid where the cheapest power now arrives and departs with the sun, and the periods between are where the price risk concentrates.

Read plainly, April 7, 2025 is a story with no disaster in it. Nobody lost power. No emergency was declared. The grid was never in danger, the weather never turned, and demand never came close to a limit. By the standards of the other events in this series, little happened. What makes it worth a story is that something did happen, and it happened entirely in the part of the grid that most people never see: a ninety-minute window in which the price of electricity across the whole of Texas went from $25 to $4,347 and back, set not by a crisis but by a maintenance calendar, a forecast that missed by 2.4 gigawatts, and the fact that the wind was not blowing in the hour before the sun came up.

That is the lesson the morning leaves behind, and it is a quieter one than a freeze or a hurricane. A modern grid does not need a catastrophe to produce a high price. It only needs its margins to thin and its forecast to slip at the same hour, and the market will find the gap and put a number on it, four thousand dollars high, before most of the state is awake. The grid can show you that number to the interval. What it cannot show you is what the number is really measuring, which is how little has to go wrong, on an ordinary spring morning, for the cost of keeping the lights on to briefly become very high.

Everything above is the same live ERCOT data behind Ask the Grid. The figures come from ercot.lmp_by_settlement_point, ercot.spp_day_ahead_hourly, ercot.system_load, ercot.wind_generation_geo, and open_meteo.era5_grid, with the ancillary-service backdrop in ercot.as_prices. Open ERCOT on the main map and scrub the morning of April 7, 2025 yourself.